Merge pull request #7585 from raffenet/4.3.x-hcoll

[4.3.x] Fix hcoll support

Approved-by: Hui Zhou <[email protected]>

Author: raffenet

src/mpid/ch4/netmod/ucx/ucx_comm.c

@@ -15,10 +15,6 @@ int MPIDI_UCX_mpi_comm_commit_pre_hook(MPIR_Comm * comm)
     int mpi_errno = MPI_SUCCESS;
     MPIR_FUNC_ENTER;
 
-#if defined HAVE_HCOLL
-    hcoll_comm_create(comm, NULL);
-#endif
-
     MPIR_FUNC_EXIT;
     return mpi_errno;
 }
@@ -29,6 +25,10 @@ int MPIDI_UCX_mpi_comm_commit_post_hook(MPIR_Comm * comm)
 
     MPIR_FUNC_ENTER;
 
+#if defined HAVE_HCOLL
+    hcoll_comm_create(comm, NULL);
+#endif
+
     MPIR_FUNC_EXIT;
     return mpi_errno;
 }

src/mpid/common/hcoll/hcoll_init.c

@@ -130,28 +130,28 @@ int hcoll_comm_create(MPIR_Comm * comm_ptr, void *param)
     int context_destroyed;
     mpi_errno = MPI_SUCCESS;
 
-    if (0 == hcoll_initialized) {
-        mpi_errno = hcoll_initialize();
-        MPIR_ERR_CHECK(mpi_errno);
-    }
+    comm_ptr->hcoll_priv.is_hcoll_init = 0;
 
     if (0 == hcoll_enable) {
-        comm_ptr->hcoll_priv.is_hcoll_init = 0;
         goto fn_exit;
     }
 
-    if (MPIR_Process.comm_world == comm_ptr) {
-        hcoll_comm_world_initialized = 1;
-    }
-    if (!hcoll_comm_world_initialized) {
-        comm_ptr->hcoll_priv.is_hcoll_init = 0;
-        goto fn_exit;
-    }
     num_ranks = comm_ptr->local_size;
     if ((MPIR_COMM_KIND__INTRACOMM != comm_ptr->comm_kind) || (2 > num_ranks)
         || comm_ptr->hierarchy_kind == MPIR_COMM_HIERARCHY_KIND__NODE_ROOTS
         || comm_ptr->hierarchy_kind == MPIR_COMM_HIERARCHY_KIND__NODE) {
-        comm_ptr->hcoll_priv.is_hcoll_init = 0;
+        goto fn_exit;
+    }
+
+    if (0 == hcoll_initialized) {
+        mpi_errno = hcoll_initialize();
+        MPIR_ERR_CHECK(mpi_errno);
+    }
+
+    if (MPIR_Process.comm_world == comm_ptr) {
+        hcoll_comm_world_initialized = 1;
+    }
+    if (!hcoll_comm_world_initialized) {
         goto fn_exit;
     }
 
@@ -194,8 +194,6 @@ int hcoll_comm_destroy(MPIR_Comm * comm_ptr, void *param)
 
 int hcoll_do_progress(int vci, int *made_progress)
 {
-    *made_progress = 1;
-
     /* hcoll_progress_fn() has been deprecated since v4.0. */
 #if HCOLL_API < HCOLL_VERSION(4,0)
     MPID_THREAD_CS_ENTER(VCI, MPIDIU_THREAD_HCOLL_MUTEX);

src/mpid/common/hcoll/hcoll_rte.c

@@ -55,7 +55,7 @@ static void progress(void)
         /* FIXME: The hcoll library needs to be updated to return
          * error codes.  The progress function pointer right now
          * expects that the function returns void. */
-        ret = hcoll_do_progress(&made_progress);
+        ret = hcoll_do_progress(-1, &made_progress);
         MPIR_Assert(ret == MPI_SUCCESS);
     }
 }